Last Class: RPCs. Today:

Size: px

Start display at page:

Download "Last Class: RPCs. Today:"

Buddy Allan Paul
5 years ago
Views:

1 Last Class: RPCs RPCs make distributed computations look like local computations Issues: Parameter passing Binding Failure handling Lecture 4, page 1 Today: Case Study: Sun RPC Lightweight RPCs Remote Method Invocation (RMI) Design issues Lecture 4, page 2

2 RPC Implementation Issues Choice of protocol [affects communication costs] Use existing protocol (UDP) or design from scratch Packet size restrictions Reliability in case of multiple packet messages Flow control Copying costs are dominant overheads Need at least 2 copies per message From client to NIC and from server NIC to server As many as 7 copies Stack in stub message buffer in stub kernel NIC medium NIC kernel stub server Scatter-gather operations can reduce overheads Lecture 4, page 3 Case Study: SUNRPC One of the most widely used RPC systems Developed for use with NFS Built on top of UDP or TCP TCP: stream is divided into records UDP: max packet size < 8912 bytes UDP: timeout plus limited number of retransmissions TCP: return error if connection is terminated by server Multiple arguments marshaled into a single structure At-least-once semantics if reply received, at-least-zero semantics if no reply. With UDP tries at-most-once Use SUN s external Data Representation (XDR) Big endian order for 32 bit integers, handle arbitrarily large data structures Lecture 4, page 4

clnt_create to locate server port Upon return, client can call procedures at the server Lecture 4, page

3 Binder: Port Mapper Server start-up: create port Server stub calls svc_register to register prog. #, version # with local port mapper Port mapper stores prog #, version #, and port Client start-up: call clnt_create to locate server port Upon return, client can call procedures at the server Lecture 4, page 5 Rpcgen: generating stubs Q_xdr.c: do XDR conversion Detailed example: later in this course Lecture 4, page 6

4 Lightweight RPCs Many RPCs occur between client and server on same machine Need to optimize RPCs for this special case => use a lightweight RPC mechanism (LRPC) Server S exports interface to remote procedures Client C on same machine imports interface OS kernel creates data structures including an argument stack shared between S and C Lecture 4, page 7 Lightweight RPCs RPC execution Push arguments onto stack Trap to kernel Kernel changes mem map of client to server address space Client thread executes procedure (OS upcall) Thread traps to kernel upon completion Kernel changes the address space back and returns control to client Called doors in Solaris Lecture 4, page 8

5 Doors Which RPC to use? - run-time bit allows stub to choose between LRPC and RPC Lecture 4, page 9 Other RPC Models Asynchronous RPC Request-reply behavior often not needed Server can reply as soon as request is received and execute procedure later Deferred-synchronous RPC Use two asynchronous RPCs Client needs a reply but can t wait for it; server sends reply via another asynchronous RPC One-way RPC Client does not even wait for an ACK from the server Limitation: reliability not guaranteed (Client does not know if procedure was executed by the server). Lecture 4, page 10

6 Asynchronous RPC 2-12 a) The interconnection between client and server in a traditional RPC b) The interaction using asynchronous RPC Lecture 4, page 11 Deferred Synchronous RPC A client and server interacting through two asynchronous RPCs 2-13 Lecture 4, page 12

7 Remote Method Invocation (RMI) RPCs applied to objects, i.e., instances of a class Class: object-oriented abstraction; module with data and operations Separation between interface and implementation Interface resides on one machine, implementation on another RMIs support system-wide object references Parameters can be object references Lecture 4, page 13 Distributed Objects When a client binds to a distributed object, load the interface ( proxy ) into client address space Proxy analogous to stubs Server stub is referred to as a skeleton Lecture 4, page 14

8 Proxies and Skeletons Proxy: client stub Maintains server ID, endpoint, object ID Sets up and tears down connection with the server [Java:] does serialization of local object parameters In practice, can be downloaded/constructed on the fly (why can t this be done for RPCs in general?) Skeleton: server stub Does deserialization and passes parameters to server and sends result to proxy Lecture 4, page 15 Binding a Client to an Object Distr_object* obj_ref; obj_ref = ; obj_ref-> do_something(); Distr_object objpref; Local_object* obj_ptr; obj_ref = ; obj_ptr = bind(obj_ref); obj_ptr -> do_something(); (a) (b) //Declare a systemwide object reference // Initialize the reference to a distributed object // Implicitly bind and invoke a method //Declare a systemwide object reference //Declare a pointer to local objects //Initialize the reference to a distributed object //Explicitly bind and obtain a pointer to the local proxy //Invoke a method on the local proxy a) (a) Example with implicit binding using only global references b) (b) Example with explicit binding using global and local references Lecture 4, page 16

9 Parameter Passing Less restrictive than RPCs. Supports system-wide object references [Java] pass local objects by value, pass remote objects by reference Lecture 4, page 17 DCE Distributed-Object Model a) Distributed dynamic objects in DCE. b) Distributed named objects Lecture 4, page 18

10 Java RMI Server Defines interface and implements interface methods Server program Creates server object and registers object with remote object registry Client Looks up server in remote object registru Uses normal method call syntax for remote methos Java tools Rmiregistry: server-side name server Rmic: uses server interface to create client and server stubs Lecture 4, page 19 Java RMI and Synchronization Java supports Monitors: synchronized objects Serializes accesses to objects How does this work for remote objects? Options: block at the client or the server Block at server Can synchronize across multiple proxies Problem: what if the client crashes while blocked? Block at proxy Need to synchronize clients at different machines Explicit distributed locking necessary Java uses proxies for blocking No protection for simultaneous access from different clients Applications need to implement distributed locking Lecture 4, page 20

Last Class: RPCs. Today:

Last Class: RPCs. Today: Last Class: RPCs RPCs make distributed computations look like local computations Issues: Parameter passing Binding Failure handling Lecture 8, page 1 Today: Lightweight RPCs Remote Method Invocation (RMI)